For the past month, since the N.F.L. banned nonessential travel and shuttered team facilities, many coaches, scouts and other player evaluators have had to rely primarily on watching footage of college games in the hopes of finding a franchise-lifting selection in the N.F.L. draft that starts on April 23.
Watching video, though, is not a replacement for meeting prospects at pro days or private workouts, where those responsible for vetting picks are able to verify a prospect’s speed, ability to separate from opponents and other statistics used to assess talent.
That’s where Slants, a start-up founded by two former college roommates who now work at Google, has tried to fill the void. The company’s artificial intelligence technology analyzes video of college football games to identify formations, routes and tracking metrics, like a player’s speed, so that coaches can better evaluate players.
“The teams don’t have any data on many college players, which is why they want to get up close at the combine and at pro days,” said Omar Ajmeri, the company’s co-founder who has been showing the technology to N.F.L. teams for three years. “After pro days were shut down, teams called and asked about our analytics. We collect tracking data not available anywhere else.”
Before the coronavirus-related travel restrictions were put into place, N.F.L. teams saw some of the 300 top draft prospects run the 40-yard dash, bench press hundreds of pounds and run and defend pass routes at the college scouting combine in late February. They also have firsthand information on many other players who completed pro days before the league halted travel. But talent evaluators are still in the blind about hundreds of other college players who played at smaller schools, are recovering from injury or who received less playing time and could prove valuable to a team searching for lesser known prospects.
In the past month, more N.F.L. teams have been sending Slants video of dozens of plays from college games and asking for predictive statistics on players — particularly wide receivers, running backs and defensive backs. Some teams are using Slants’s findings to confirm data that their scouts collected in games or drills. In other cases, Slants is providing new information.
“It kind of feels guilty to say with everything going on in the world, but it’s helped show people the value of this technology,” said Ali Shah, the company’s other co-founder.
N.F.L. teams are reluctant to openly discuss how they use the technology for fear of giving their rivals an advantage. But one person in charge of analytics at an N.F.L. team who spoke on condition of anonymity said that he is using Slants to help evaluate some of the 3,000 or so college players he has tracked ahead of this year’s draft. His team also uses Pro Football Focus, which grades college players from the Power 5 conferences. His staff also watches video to subjectively evaluate lesser known players.
“But being in analytics, we think there’s a more scientific approach,” the team executive said. Slants’s video technology can estimate a player’s 40-yard speed, which “is incredibly useful to teams because speed and yards of separation are influential and predictive variables.”
The first iterations of Slants’s software were developed three years ago by Ajmeri, who grew up a Washington Redskins fan in Maryland and dreamed of becoming an N.F.L. general manager. He talked to Shah, his college friend, about how help football teams become more efficient. The duo, who also follow basketball and soccer, believe football is not as advanced as other sports in using automated technology.
“We had maybe five ideas of where football is lagging behind, where it feels more old school than new school,” Shah said.
Ajmeri worked as an intern at the N.F.L. in the summer of 2013 and had a feel for how the league worked, and Shah, a computer programmer who roots for the Jets, helped translate their ideas into software.
Their first public validation came in 2018, when they presented their technology at the influential M.I.T. Sloan Sports Analytics Conference. A panel of outside judges gave them an “Alpha Award” for best research paper. At the conference, they met representatives from N.F.L. teams who were looking to automate the often ponderous task of tagging game footage. Several teams continued to work with Slants and provide them with tips on what N.F.L. teams need, including ways to evaluate college players.
Ajmeri and Shah sought to replicate the statistics that N.F.L. teams collect from their own players, who all wear radio-frequency identification (or R.F.I.D.) chips in their shoulder pads so that antennas and beacons in stadiums can track their movements on the field. Colleges do not use this technology yet, another reason pro teams rely so heavily on collecting their own data to size up players.
“At the end of the day, we have the N.F.L. data and can process it, but there’s nothing like it on the college level,” said Tyler Oberly, who worked with Slants when he ran the Tampa Bay Buccaneers’ analytics department from 2014 to 2019.
The big leap for Slants came in mid-2018, when Omar’s younger brother, Ameen, joined and used his skills in machine learning to teach the software to compensate for different camera angles and other variables.
The three men continue to hold day jobs and meet at Omar and Ameen’s apartment in Brooklyn at night to iron out their technology and meet requests from N.F.L. teams. Slants has no revenue now, but they are hoping that after the pandemic subsides, they will be able to sign contracts to work with teams throughout the year.
They join an increasingly crowded field of companies using tracking technology to help teams evaluate player performance. Major League Baseball has for years tracked pitch speeds and types, as well as player defense. N.B.A. arenas house cameras trained on game action as part of the Second Spectrum Player Tracking System, recording data about players, the referees and the ball.
“The general class of technology has been out for years and the N.B.A. has been using it the most,” said Daryl Morey, the general manager of the Houston Rockets who helps organize the M.I.T. Sloan conference. “But football is a natural fit for it because you have data from an overhead camera that you want to understand.”
Slants’s data collecting is still being refined. The company does not use expensive footage from TV broadcasts. Rather, it analyzes video from the so-called All 22 camera angle above the 50-yard line taken by college football assistants. The horizontal view of the field can make it hard for the software to identify jersey numbers and big offensive linemen can sometimes obscure players at certain positions. The software had to be taught to ignore distractions like referees, coaches and players on the sidelines and fans in the stands.
The software’s mapping algorithm improves as it watches more video. Slants’s technology can gauge a player’s speed to within half-a-mile per hour, based on comparisons with publicly available next-generation statistics that the N.F.L. releases for its players, Ajmeri said.
Unlike other tracking technology, Slants does not require any equipment in stadiums and it can analyze game footage from years ago, before the N.F.L. began tracking players.
Oberly, who now works outside of sports, said Slants’s technology could also be used to analyze players in high school or playing overseas, adding that the company has a head start on potential rivals seeking to replicate the technology for other sports like soccer or hockey. “It’s an entirely new set of problems,” he said.